Fitting ("Estimating") Multivariate Normal Distribution to Data

Hello,

I would like to fit a multivariate normal distribution to a few variables in Matlab. The fitted distribution would then be used to generate simulated data in a Monte Carlo exercise. I'm just wondering what is the advantage of using the "fitgmdist" command over just estimating the mean with "mean" and the variance-covariance matrix with "cov". I'm new to "fitdist" and am unfamiliar with the seemingly dizzling array of options.

Many Thanks.

Answer 1

[Dheeraj Singh]

fitgmdist provides you different options which you can vary depending upon your data to get better results.

You can adjust the number of iterations for the EM algorithm if you want your algorithm to run within some time.

If you have less data, you can adjust the ‘RegularizationValue’ to avoid overfitting.

There are many options which we can use if our solution does not converge.

For more detail, you can refer to the following link:

https://in.mathworks.com/help/stats/fitgmdist.html.